Scientific Machine Learning for Faster Simulation and Physical Insight

We combine physical models with machine learning to reduce computational cost and understand complex systems. In cardiovascular modeling, we develop fast approximations of patient-specific blood flow that retain a physical structure while learning corrections from detailed simulations.

For expensive engineering studies, our active-learning methods select which simulations to run and at what level of accuracy to reduce uncertainty efficiently.

We also develop interpretable representations and generative models that reveal distinct physical regimes in fluid and combustion data. Together, these methods support faster prediction, exploration of design choices, and analysis of physical behavior.